Matrix Effect Correction And Application Of Portable X-ray Fluorescence Spectrometry

Portable X-ray fluorescence spectrometry(pXRF)is a low-cost test method capable of performing multi-element,rapid,and non-destructive analysis in the field,making it ideal for difficult-to-obtain samples,valuable undamaged samples,or largescale samples.However,the biggest problem in its application is that the matrix effect between elements during the testing process has a significant impact on the quality of the pXRF data,and the correction of the matrix effect at this stage is essentially the linear or nonlinear regression model,which is based on the absolute residual sum of squares minimum principle.But the flexibility and universality of these models are greatly limited in the case of complicated and changeable sample matrices.In this study,a new pXRF matrix effect correction model was proposed that merged the linear regression and the nonlinear regression while balancing the absolute residual sum of squares minimum principle and the relative residual sum of squares minimum principle.And the correlations between models and elements were evaluated using novel metaanalysis,correspondence analysis and traditional statistical parameters.In order to verify the applicability and generality of the new model,it was then applied in several research fields to assess its calibrationimpact.The main results obtained in this study are as follows:1.A new pXRF matrix effect correction model was developed that is dynamic,adaptable,and blends data type with user judgment.The model chose 7 major elements(Al,Si,K,Ca,Ti,Mn,and Fe)that were more stable and had a high content to replace the sample matrix,weakening the barrier to matrix effect correction due to the large number and complexity of components in the matrix.The model combined the absolute residual sum of squares minimum principle and the relative residual sum of squares minimum principle were combined,the former was responsible for the high correction and the latter for the low correction.The combination of the two solved the issue of inhomogeneous matrix effects in the presence of high concentration discrepancies.Meanwhile,the model integrated linear regression and nonlinear regression,with the former being in charge of stability and overfitting and the latter of accuracy and underfitting.The combination of the two improved the accuracy and precision of the model.As a result,the new model was a dynamic cubic model rather than a single fixed model.2.Compared with the previous model,the new model had a stronger correction effect for elements,and the increase in data quality was greater,with an average improvement of over 90% and 70% in absolute and relative degrees,respectively.Especially in the relative degree,the new model safeguarded the low content and could still correct for matrix effects even if the overall content of elements was considerably different,which was the major difference between the new model and the previous model.The new model had a good correction effect for the 34 elements selected,with Mg,V,Cr,Co,Mo,Cd,Sn,Ba,W,and U having the best correction effect.3.Multiple statistical methods were used for a thorough evaluation of the models.Meta-analysis in medical fields was performed to compare models overall,and outstanding application results were produced,clearly demonstrating the degree of improvement in model correction impact.Instead of traditional factor analysis,correspondence analysis was used to determine the relationship between models and elements,which summarized their applicability.Traditional statistical parameters were used to assess the benefits and drawbacks of models.In the construction of the new model,each statistical method yielded good results.4.The new pXRF matrix effect correction model was applied in 4 research fields,which were the polymetallic mining area,the precious metal mining area,the heavy metal pollution of soils,and rock weathering.The application results were relatively positive.Comparing the results of the study with those of other studies,the majority of them were more consistent with less inaccuracy,which confirmed the credibility and generality of the model.Some problems in the study area were investigated using calibrated data,and suggestions were made based on the results,including the location of mineralization zones of deposits,the prediction of burial of blind ore bodies,the status and distribution of heavy metal contamination,and the applicability of weathering indices to different lithologies.5.The new pXRF matrix effect correction model produced varying results in different fields,with the Asher copper-zinc deposit being the most effective,followed by the Qujia gold deposit and karst rock weathering,and finally heavy metal pollution.The higher the target element content in the sample,the higher the detection ratio,the greater the completeness of the data,the easier the model finds the intrinsic pattern,the less difficult the matrix effect correction,and the higher the data quality.The application was also influenced by the morphology of the medium,including its compactness and homogeneity,which influenced the chemical matrix interaction through the morphology of the physical matrix.In summary,the new pXRF matrix effect correction model was obtained by combining the absolute residual sum of squares minimum principle with the relative residual sum of squares minimum principle and the linear regression with the nonlinear regression.As compared to the previous model,the new model had a better correction effect on pXRF data and greater improvement on data quality,with an average improvement of over 90% and 70% in the absolute and relative degree,as well as more stability and applicability.And the new model had been successfully applied in several research fields such as deposit exploration,pollution monitoring,and rock weathering,demonstrating the reliability and universality of the model.